Skip to content
Notes by Axbom
Per Axbom LinkedIn

Some tasks that are being handed over to automation should arguably never have been tasks in the first place. Or should at least be revised regularly. But instead of questioning and reinventing tasks, many will now be automated to run perpetually. And with no human involved, why would someone then question its validity?

I believe this is part of the non-intuitive aspect of AI and automation, that it will actually in many instances resist change rather than encourage it.

People will forget why that machine is running and worry that cutting its power will break something.

When, in fact, it might *free* something.

#Automation #AiEthics #DigitalEthics

🔗 Originally posted on LinkedIn

30 Jun 2023, 10:22
Share
LinkedIn Bluesky Email
Per Axbom Akkoma

Some tasks that are being handed over to automation should arguably never have been tasks in the first place. Or should at least be revised regularly. But instead of questioning and reinventing tasks, many will now be automated to run perpetually. And with no human involved, why would someone then question its validity?

I believe this is part of the non-intuitive aspect of AI and automation, that it will actually in many instances resist change rather than encourage it.

People will forget why that machine is running and worry that cutting its power will break something.

When, in fact, it might *free* something.

https://axbom.com/automation-change/

#AiEthics #DigitalEthics

Your regular reminder that tech should work FOR YOU, sure, but *also* not against those who are already underserved.

A line drawing. Two people talking about a black box on the floor, connected to an electronic display on the wall.

Person 1: What does this box do?
Person 2: Dunno, it's been there for a decade.
Person 1: Better not touch it then.

🔗 Originally posted on Akkoma

30 Jun 2023, 09:23
Share
LinkedIn Bluesky Email
Per Axbom Akkoma

En del svÄrsmÀlta sommarprat hittills...

Av de fem jag lyssnat pÄ hittills gillar jag Jessie Sommarströms mest, tÀtt följd av Axel Gordh Humlesjö - som var nÄgot av det svÄraste jag lyssnat pÄ.

Jag gör en ansats att lyssna pÄ alla och betygsÀtta för mig sjÀlv. HjÀlper mig att förstÄ vad jag vÀrderar. Och kanske vÄgar jag utforska varför.

#sommar #SommariP1

🔗 Originally posted on Akkoma

30 Jun 2023, 08:42
Share
LinkedIn Bluesky Email
Per Axbom Akkoma

FYI. If you have a micro.blog account you can import your Twitter archive there, giving you online access to all your tweets (videos not included).

https://help.micro.blog/t/importing-tweets/1677

🔗 Originally posted on Akkoma

29 Jun 2023, 15:48
Share
LinkedIn Bluesky Email
Per Axbom Akkoma reply

@m I've been using Raindrop for a few years now. Absolutely love it.

It's also how I share my digital ethics links on a public page:
https://raindrop.io/axbom/digital-ethics-6334413

🔗 Originally posted on Akkoma

29 Jun 2023, 14:20
Share
LinkedIn Bluesky Email
Per Axbom LinkedIn

Bra debattartikel om AI-etik i vÄrden. Det Àr det omrÄde dÀr jag ser enorm potential för att rÀdda liv, men ocksÄ oerhörda risker med obalans i vems liv som prioriteras. AnsvarsfrÄgan blir fruktansvÀrt viktig att reda ut innan systemen tas i bruk.

David Sundemo och Sam Polesie lyfter ansvarsfrÄgan pÄ ett tydligt sÀtt:

"AnsvarsfrĂ„gan kan exemplifieras pĂ„ följande sĂ€tt; Rut genomgĂ„r en röntgenundersökning. Ett AI-system bedömer att det inte finns nĂ„got avvikande och Rut fĂ„r lugnande besked. Ett Ă„r senare söker hon vĂ„rd igen och det visar sig att hon har en tumör i höger lunga. Vid eftergranskning framgĂ„r att tumören kunde ses redan pĂ„ den första bilden. Vem Ă€r ansvarig? Är det ingenjörerna som utvecklade algoritmen, Ă€r det den som köpte in systemet, eller Ă€r det lĂ€karen som förmedlat beskedet?"

Hur dessa system trÀnas blir ocksÄ en nyckelfrÄga. DÄ vita mÀns hÀlsa och kroppar lÀnge har fÄtt stÄ modell för sjukvÄrdens forskning sÄ Àr det ocksÄ denna data som mÄnga AI-system matas med. Det blir dÄ en underrepresentation av trÀningsdata för personer som ofta utstÄr diskriminering i andra sammanhang ocksÄ.

"En lÀkare som exempelvis bara trÀnar pÄ att bedöma hudförÀndringar hos ljushyade blir duktig pÄ det, men mindre bra pÄ att bedöma förÀndringar hos mörkhyade. I databaser för trÀning av hudbedömningsalgoritmer Àr personer med mörk hud kraftigt underrepresenterade. Om detta fortsÀtter Àr risken stor att hÀlsomÀssiga ojÀmlikheter cementeras."

Jag brukar uttrycka det sÄ hÀr:

Hur ska en dators vĂ€rderingar förĂ€ndras nĂ€r mĂ€nniskors vĂ€rderingar förĂ€ndras över tid? Modellerna vilar nĂ€mligen ofta pĂ„ decennier av fördomsfullt informationsunderlag – och oerhört begrĂ€nsade underlag ur ett globalt perspektiv. Att förflytta beslutsfattande till maskiner riskerar alltsĂ„ att öka beslutsfattande som Ă€r grundat i vĂ€rderingar mĂ„nga mĂ€nniskor vill komma ifrĂ„n.

Till sist poÀngterar Sundemo och Polesie oksÄ hur lÀkare mÄste vara mer involverade i utvecklingen av verktygen. Vi riskerar annars att patienter reduceras till datapunkter som behandlas som statistiska fenomen.

Jag vill nog pÄstÄ att patienter och vÄrdsökande mÄste vara mer involverade ocksÄ. Det Àr de som alla andras makt mest tydligt drabbar. Och dÄ gÀrna de som har sÀrskild erfarenhet av att diskrimineras och bli bortglömd i alla tekniksprÄng som ska förbÀttra vÀrlden för alla, men ofta bara gör det för ett sÀrskilt urval.

https://lakartidningen.se/opinion/debatt/2022/08/ai-etik-en-nyckelfraga-for-etiskt-hallbar-vard/

#AiEtik #DigitalEtik #DigitalOmtanke

🔗 Originally posted on LinkedIn

29 Jun 2023, 09:20
Share
LinkedIn Bluesky Email
Per Axbom LinkedIn

Is there a controlled vocabulary for AI governance? Why yes, yes there is. đŸ”„

🔗 Originally posted on LinkedIn

29 Jun 2023, 07:06
Share
LinkedIn Bluesky Email
Per Axbom Akkoma reply

@m

Exactly. I know this is a sensitive topic in the design industry, but as I’ve always coded I know and appreciate the huge benefits of something that behaves like the final product.

And I don’t think all designers should be able to do this (but to at least understand the material they are working with), but my feeling is that the deep feature set of FIgma is becoming more akin to coding than designing. Which can be alright if it creates value that is not better achieved another way, it’s just a different type of profession. And I’m an old geezer trying to make sense of the world. 😅

🔗 Originally posted on Akkoma

29 Jun 2023, 06:46
Share
LinkedIn Bluesky Email
Per Axbom LinkedIn

Really good, concise segment on the US perspective of regulating AI. And the type of AI that we should be encouraging and supporting. 👏

For more AI wisdom you should be following Dr. Joy Buolamwini and signing up for her newsletter. For those not familiar with her work she is based at MIT Media Lab and founder of The Algorithmic Justice League

🔗 Originally posted on LinkedIn

29 Jun 2023, 06:36
Share
LinkedIn Bluesky Email
Per Axbom Akkoma reply

@esbenrahbek Tack Esben!

🔗 Originally posted on Akkoma

28 Jun 2023, 22:02
Share
LinkedIn Bluesky Email
Per Axbom Akkoma reply

@simeon This is fair. I think I'm mostly concerned about how much effort is put into high fidelity. I often do rough sketches with a pen, move stuff around, test, get feedback. Combinations of paper, digital whiteboards and Keynote (quirky I know, but also extremely fast). Then I head to HTML :)

I feel the learning curve for all of Figma's features is similar to the learning curve of learning HTML and CSS, but I do acknowledge there are collaboration benefits.

I know my process is far from "standard" 😅

🔗 Originally posted on Akkoma

28 Jun 2023, 20:24
Share
LinkedIn Bluesky Email
Per Axbom Akkoma

As I'm reading about the ever-increasing complexity and feature-set of Figma I can't help but wonder if it wouldn't be easier if designers just learned HTML and CSS. There seems to be an awful lot of work going into creating the likeness of something (a web page) without it being that thing.

I am likely too grumpy about this.

🔗 Originally posted on Akkoma

28 Jun 2023, 20:03
Share
LinkedIn Bluesky Email
Per Axbom LinkedIn

So much good content that could've been published as HTML (a web page) is being published as PDF documents. 😭

Often great reports on important topics are published in 2-column PDFs which are just impossibly awkward to zoom and find your way in. They also suffer from low accessibility, taking up space on your computer (and in email!) and just generally being harder to find.

Now imagine that same report as a web page, nicely formatted with headings, a screen-friendly font and a width that adjusts to the phone or computer browser. And it works with screen readers and magnifying tools. You can set it up so you can send a link directly to specific sections. For extra points, you can offer dark mode and high-contrast mode to help even more people read your findings.

Honestly, let me help you make your report into a web page if you're stuck in a PDF mindset. Not really my day job, but wouldn't it be great with a service set up to help with something like this...

Image description:

Close-up of hands navigating a phone lying flat on a notebook on a table next to a coffee cup and a pen. The screen is displaying a black-and-white two-column PDF where you can see one column and part of the second column.

Title on image: Your report as a PDF – Common issues

Text on image:

- Awkward to navigate

- Low accessibility

- Fills up inboxes and hard drives perpetually

- Gets lost and is hard to find again

🔗 Originally posted on LinkedIn

28 Jun 2023, 18:46
Share
LinkedIn Bluesky Email
Per Axbom Akkoma

Updated the "If a hammer was like AI"-poster with a better font and some tweaks to topic texts to make them less ambiguous.

The font is Atkinson Hyperlegible, a free font designed for greater legibility and readability for low vision readers

PDF and PNG images available:
https://axbom.com/hammer-ai

#AIEthics #DigitalEthics

Poster: If a hammer was like AI.

Illustration of a hammer (a line drawing) and the quote

🔗 Originally posted on Akkoma

28 Jun 2023, 15:32
Share
LinkedIn Bluesky Email
Per Axbom LinkedIn

When you make one part of a system more efficient, or even friction-free, other parts of the system may wear out and break. What do I mean by this?

Making only one cog in your machine more efficient will put unexpected pressure on the rest of the organisation. And this can and will apply to phenomenons outside your organisation, albeit within the system you are interacting with. When external functions are taken by surprise and struggle to cope with added effort, the output of the system you were trying to improve may instead deteriorate.

I've been reading news about the effects of the crash detection feature implemented in Apple's phones and watches, and likely present in devices by other manufacturers as well. It's been shown how often these detection systems falsely trigger a call to emergency services, increasing the load on operators who are trained to take every call seriously. The number of "silent calls" to emergency services in Sweden (112) has increased by 40% in just one year. And last week the 112 service released data from an investigation saying that long wait times _may_ have contributed to 5 deaths last summer.

There is no way of determining the phone feature was a contributor to this finding. But it certainly wasn't of help to the stability of the system. And should raise alarms for continued operations with no maintenance efforts.

Driving innovation through compassion for me means:

* seeking counsel with the organisations/people you are targeting and affecting,

* applying strategic foresight to identify these types of issues before they happen,

* monitoring and taking responsibility for – and being prepared to walk back – ideas that didn't work out,

* working for systems where cogs are monitored and built for harmonious movement rather than disconnect.

Yes, we can find cases where these features have saved lives. But it's unclear at what cost this change is happening, or even how crucial this exact feature was in those cases. It's likely relevant to consider if the ownership of a specific phone should be what determines who receives attention.

Likely there are many ways to address this issue. Meanwhile, accountability is largely missing for the changing work environment of emergency call operators.

IMAGE DESCRIPTION:

Three aligned, interconnected cogs with labels.

Cog number 1 (in good condition).

Label: Phones and watches auto-calling emergency services

News clipping: Skiing triggers crash detection on latest Apple Watch and iPhone

Cog number 2 (showing wear and tear).

Label: Increase in "silent calls" due to false emergencies

News clipping: Huge increase in misdials to emergency services

Detail: Up 40% in Sweden in one year.

Cog number 3 (showing wear and tear).

Label: Longer wait times for people in real emergencies

News clipping: 5 people died waiting several minutes to reach emergency services

#DigitalCompassion #SystemsThinking #ResponsibleInnovation

🔗 Originally posted on LinkedIn

28 Jun 2023, 14:56
Share
LinkedIn Bluesky Email
Per Axbom Akkoma reply

@steveportigal Oh wow I had no idea they were shutting down.

🔗 Originally posted on Akkoma

27 Jun 2023, 18:49
Share
LinkedIn Bluesky Email
Per Axbom LinkedIn

Published earlier today: Deepfake porn using any person on Instagram

When we can no longer control where our face appears, what happens to our perception of identity? The technology is constantly getting better in its deceptive quality, is more widely available and at ever lower costs. The porn industry is growing in all directions. As it stands, today everyone is a potential porn star against their own will.

https://axbom.com/deepfake-porn/

🔗 Originally posted on LinkedIn

26 Jun 2023, 21:59
Share
LinkedIn Bluesky Email
Per Axbom Akkoma reply

@ikt

Yeah, sorry about that. I honestly have a hard time writing about stuff like this but I feel like it will only be worse if it's ignored and/or not talked about.

🔗 Originally posted on Akkoma

26 Jun 2023, 17:00
Share
LinkedIn Bluesky Email
Per Axbom Akkoma

Just published: Deepfake porn using any person on Instagram

When we can no longer control where our face appears, what happens to our perception of identity? The technology is constantly getting better in its deceptive quality, is more widely available and at ever lower costs. The porn industry is growing in all directions. As it stands, today everyone is a potential porn star against their own will.

https://axbom.com/deepfake-porn/

#AiEthics #DigitalEthics

🔗 Originally posted on Akkoma

26 Jun 2023, 10:31
Share
LinkedIn Bluesky Email
Per Axbom LinkedIn

”By simply existing as women in public life, we have all become targets, stripped of our accomplishments, our intellect, and our activism and reduced to sex objects for the pleasure of millions of anonymous eyes.”

My previous article in Swedish on deepfake porn can be found here:

https://axbom.se/deepfake-porr/

I’ve added a link to Nina Jankowiscz’s article there as well.

🔗 Originally posted on LinkedIn

25 Jun 2023, 21:01
Share
LinkedIn Bluesky Email
← Newer Older →
  • Notes by Axbom
  • RSS

© 2026 Notes by Axbom